摘要
Apriori算法自身虽然进行了一定的优化,但它无法对最庞大的2阶候选项集进行削减,频繁项集中也至少有80%的项集不包含有效规则。其它改进算法虽然从不同角度对原算法进行了优化,但也不能解决后一个问题。文中所研究的算法,在保留原算法优化的基础上,首先引入隐含度概念及隐含度削减算法,对庞大的2阶候选项集进行削减;其次利用置信度,对k≥2阶频繁项集进行削减,同步生成关联规则,从而提高算法效率;最后讨论了只挖掘单项关联规则的可行性,仅需扫描原始数据库2次。
Although Apriori algorithm itself gets optimization at some extent,it has no effect on the enormous set of candidate 2-itemsets,and can not avoid more than 80 percent of itemsets which involved no expected rules.Meanwhile other optimized algorithms improved it from different aspects,they could not avoid the latter as well.Discussed two methods to improve the efficiency of Apriori.Firstly,introduces the concept of implicit and uses it to reduce the size of the set of candidate 2-itemsets.Secondly,uses confidence to generate association rules and removes the frequent itemsets which does not involve any expected rule,so as to reduce the size of the set of frequent itemsets.At last,argued generating the patterns of association rules using 2-itemsets,without creating any upper itemsets.It needs to scan the original transaction database only twice.
出处
《计算机技术与发展》
2010年第11期105-108,共4页
Computer Technology and Development
基金
贵州省科技计划攻关课题项目(黔科合GY字[2008]3035)